Seeing if Hex-Rays' decompiler plugin (http://www.hex-rays.com/decompiler.shtml) for IDA Pro (http://www.hex-rays.com/idapro) can be reimplemented in Python with little work for much benefit, following the Pareto Principle. Not really cloning their work; rather, noting that IDA does not expose even trivial data-flow analysis.

I suspect that most of the fancy stuff is locked up for decompiler users only, which would make sense since the decompiler is much more expensive than IDA itself. So far, I have du and ud chains implemented in a nice, readable fashion and will be releasing the codebase this week, after my advisor and I discuss licensing issues. Going to write well-engineered code for automated analyses that are often done in an ad-hoc fashion by the RCE community.

EDIT: after reading another comment about how short news.yc is, just shortened an idiom that was repeated 5 or 6 times. It would be sort of nice if this was done at compile time to avoid the overhead of function calls and nested generators, but it's still quite elegant IMO to clean up common loops with generators.